Bug 68969 - xf86-video-intel 2.99.901 + XMir + multimonitors = all displays black
Summary: xf86-video-intel 2.99.901 + XMir + multimonitors = all displays black
Status: RESOLVED NOTOURBUG
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: git
Hardware: Other All
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-09-05 09:35 UTC by Daniel van Vugt
Modified: 2013-09-06 08:17 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments

Description Daniel van Vugt 2013-09-05 09:35:36 UTC
xf86-video-intel 2.99.901 + XMir + multimonitors = all displays black

With a secondary monitor plugged in, the latest xf86-video-intel 2.99.901 with XMir results in all black displays.

Using only a single monitor there is no problem. It works fine.
Comment 1 Chris Wilson 2013-09-05 11:38:52 UTC
Bug in XMir (xmir_resize() is fubar), and Mir itself is broken wrt to pageflips (see https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/1220458)
Comment 2 Daniel van Vugt 2013-09-06 01:51:38 UTC
Hmm OK. It all works fine with Ubuntu's xserver-xorg-video-intel 2:2.21.14-4ubuntu3. I assumed therefore that the XMir package (unchanged) would not be to blame.
Comment 3 Daniel van Vugt 2013-09-06 06:02:12 UTC
Chris, can you please provide more info?...

(a) "xmir_resize() is fubar"; how? So that we might make it less "fu"'d.

(b) "Mir itself is broken wrt to pageflips"; Please elaborate. In the LP bug even.

(c) If 2:2.21.14-4ubuntu3 works, why is it not a bug that 2.99.901 fails?
Comment 4 Chris Wilson 2013-09-06 08:09:21 UTC
(In reply to comment #3)
> Chris, can you please provide more info?...
> 
> (a) "xmir_resize() is fubar"; how? So that we might make it less "fu"'d.

Aside from it being a hack around broken drivers, it releases a reference it did not own.

> (b) "Mir itself is broken wrt to pageflips"; Please elaborate. In the LP bug
> even.

The dmesg is full of incorrect ioctl calls. Then Mir promptly dies after being told that it did something wrong. And then the races truely begin.

> (c) If 2:2.21.14-4ubuntu3 works, why is it not a bug that 2.99.901 fails?

Your code is broken. If you looked at the traces, you would see that it can't possibly be doing what you expected. Mir is showing buffers out of sequence with XMir (during normal usage), and after the modeset XMir is not receiving the new fd, but continues to render to the previous pair.
Comment 5 Daniel van Vugt 2013-09-06 08:17:32 UTC
(a) We shall have to investigate that... RAOF?

(b) I'm happy to fix any incorrect ioctls (made via the drm wrapper library presumably). However it's not immediately obvious what's wrong if the driver continues to report success back to Mir.

(c) I'm working on an out-of-order problem right now. It appears XMir and xf86-video-intel are getting out of sequence (not Mir itself) so any help would be appreciated:
https://bugs.launchpad.net/xmir/+bug/1216472


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.